Optimal design and operation of an Organic Rankine Cycle (ORC) system driven by solar energy with sensible thermal energy storage

In this study, the optimal design and operation of an Organic Rankine Cycle (ORC) system driven by solar energy is investigated. A two-tank sensible thermal energy storage system is configured to overcome the intermittency of solar energy. A circulating fluid, also termed as heat transfer fluid (HTF) that connects the solar collector and the ORC system plays a critical role in this system. The mass flowrate of the HTF determines both the temperature of the HTF and the amount of heat absorbed from the solar collector. A simulation-based optimization model is developed in this work. Process simulation of the ORC is performed in Aspen HYSYS, and the mathematical models of the energy storage system and the parabolic trough collector are developed in Matlab. The optimal design of the system including the hot tank temperature, cold tank temperature, mass flowrate of the HTF, and operating conditions of the ORC are determined simultaneously based on the simulation-based optimization framework. The control strategy of the solar collector can be determined as well. The system efficiency of the solar energy driven ORC system is maximized with the proposed optimal operation strategy. With the simulation-based optimization framework, the system efficiency of the recuperative ORC power plant with toluene as the working fluid is increased from 17.9% to 24.8% compared with a previous study in the literature. The recu-perative ORC performs much better than the basic ORC. Toluene performs best among all the investigated working fluids ignoring the problem with vacuum condensation. The cycle type (subcritical vs. supercritical) exerts great influence on the system performance. The supercritical ORC can improve the thermal efficiency by 11.3% and the overall system efficiency by 10.8% compared with the subcritical ORC with n-pentane as the working fluid.


Introduction
Pressing climate change along with technological advancements has spurred the growth of renewable energy installations in the last decade [1]. The greenhouse gas emissions from conventional power generation are receiving increasing attention. Environmentally friendly and energyefficient power generation technologies represent a general trend in the future energy market. Solar energy utilization is a promising way to generate electricity because of its abundance and availability [2]. Electricity production from solar energy has been proven to be a viable option for green energy production. Among various solar energy technologies, concentrated solar power (CSP) is attractive due to its high efficiency, low operating cost, and flexible scale-up potential [3].
Furthermore, there are different CSP technologies such as parabolic trough collectors (PTC), linear Fresnel reflectors (LFR), solar power towers (SPT), and parabolic dish reflectors [4]. The PTC technology is one of the most advanced solar thermal energy technologies with considerable operational experience and has the advantage of low installation cost compared to other technologies. PTCs can effectively produce heat at temperatures between 50 • C and 400 • C [4]. Several different potential power cycles can convert heat into power. The most versatile and efficient power cycle below 400 • C is the Organic Rankine Cycle (ORC) [5], which is widely used for low-temperature waste heat recovery in the industry [6]. Although the conventional steam Rankine cycle dominates in terms of efficiency when utilizing heat sources at temperatures of 400 • C or higher, these temperatures are out of reach for many conventional PTC technologies. In addition, the ORC is more compact and less costly compared with the conventional steam Rankine cycle power plant [7].
A transcritical CO 2 cycle is also an alternative for solar energy utilization if a low temperature heat sink is available. Mehrpooya and Sharifzadeh [8] proposed a novel oxy-fuel transcritical Rankine cycle with carbon capture for the simultaneous utilization of solar energy and liquefied natural gas (LNG) cold energy. A thermal energy storage tank was adopted to overcome the intermittency of solar energy. Ahmadi et al. [9] performed a techno-economic analysis and multi-objective optimization of a transcritical CO 2 power cycle driven by solar energy and LNG cold energy. A heat storage tank and an auxiliary heater are configured in their system as well. It is obvious that energy storage is necessary for solar energy utilization by means of power cycles. However, without a perfect heat sink like LNG, an Organic Rankine Cycle is more suitable for solar energy utilization. Cocco and Serra [10] compared the solar-powered ORC system with a two-tank direct energy storage system and a thermocline energy storage system, and concluded that the two-tank direct energy storage system shows slightly higher efficiency. Wang et al. [11] investigated the off-design performance of a solar-powered ORC, where only one tank is considered. To avoid offdesign operation of the ORC system, Yang et al. [12] proposed a novel operating mode of the ORC system with a two-tank energy storage system under nominal design conditions. This novel operating mode can maintain high efficiency of the ORC system with stable power output. However, process optimization and optimal operation of the energy storage system were not thoroughly addressed in their study. Table 1 summarizes the key contributions and characteristics of relevant previous studies and this study.
The main objective of this study is to determine the optimal design and optimal operating strategy for the two-tank storage solar energy driven ORC system proposed by Yang et al. [12] and carry out a thorough performance analysis under different assumptions. The main contributions of this study are the following: (1) A simulation-based optimization framework is developed to optimize the performance of the solar driven ORC system with round-the-clock electricity generation.
(2) The impacts on the system performance of (i) system configuration (basic ORC vs. recuperative ORC), (ii) power cycle type (subcritical vs. supercritical), and (iii) working fluids are investigated thoroughly. The rest of this paper is organized as follows: Section 2 presents the detailed description of the solar energy driven ORC power plant with energy storage. The mathematical model and process simulation model along with the optimization algorithm used in this study are presented in Section 3. The optimal results and the main findings are analyzed in Section 4. Section 5 summarizes the conclusions of this study.

Process description
The peak hours of solar irradiation are usually in hours of low electricity demand in the grid. Without energy storage technology, the solar power plant has to be operated at off-design conditions most of the time, which leads to the quite low thermal efficiency of the ORC and low overall system efficiency. Therefore, thermal energy storage (TES) is normally configured to overcome this challenge, and thus the profit of the solar energy power plant can be improved [13]. With the TES system, the solar power plant can not only generate electricity round-theclock but also operate at nominal design conditions to maintain high efficiency. TES technology, acting like a buffer between the solar collector and the power generating unit, allows for flexibility of the power system. The periods with high solar intensity do not correspond to the periods with high power demand. Thus, the energy can be stored in periods with high solar intensity and low power demand, and be released in periods with low solar intensity and high power demand [3]. TES systems can be classified into sensible heat, latent heat and thermochemical energy storage technologies [14], while TES systems can also be classified as direct and indirect based on the heat transfer fluid and the storage medium [15]. The two-tank energy storage system proposed by Yang et al. [12] is shown in Fig. 1. This system has been proven to be an efficient system configuration for solar energy Yes Yes utilization with stable power output. In our study, the two-tank energy storage system, which can also be categorized as a direct and sensible thermal energy storage, is chosen as the TES system. In this study, the integrated system consists of a solar energy collecting sub-system, thermal energy storage sub-system, and an ORC power generation sub-system. The parabolic trough collector (PTC) was selected as the solar collector as it can heat the heat transfer fluid (HTF) to relatively high temperatures with good efficiency, and this technology has reached the highest level of commercial maturity. PTC systems account for the largest share of the current concentrated solar power market compared with other technologies [16]. As shown by the dashed lines in Fig. 1, by adjusting the mass flowrate of the HTF through the solar collector (m c ), the HTF can be heated to a constant temperature even under varying solar insolation. The HTF can be released at a constant mass flowrate to the evaporator, where it delivers heat to the ORC sub-system; thus, the ORC can always operate stable at nominal design conditions. During solar peak hours, the HTF travels through the PTC from the cold tank at a high flowrate and then is stored in the hot tank. Stable operation of the system can avoid a drastic decrease of system efficiency during unstable operation, as shown in the work of Wang et al. [11]. The solid lines in Figs. 1 and 2 are meant to indicate that the flowrates of these streams are constant, while the dashed lines mean that the flowrate of such streams varies with the insolation. The pumps to circulate the HTF between the hot tank and the cold tank are omitted because their work requirements are negligible compared to the power output of the ORC sub-system.
The ORC sub-system consists of a pump, an evaporator, a turbine and a condenser as shown in Fig. 1. The organic working fluid is pumped from condensation pressure to evaporation pressure (process 1 → 2). After pumping, the organic working fluid is vaporized and superheated in the evaporator (process 2 → 3). Next, the high temperature and high pressure vapor is expanded through the turbine to generate power (process 3 → 4). Finally, the working fluid is condensed in the condenser (process 4 → 1). To improve the thermal efficiency of the ORC subsystem, a recuperator can be configured between the turbine outlet stream and the pump outlet stream to recover part of the condensation heat of the organic working fluid. Fig. 2 illustrates the layout of the integrated system with a recuperator.
With this configuration, the HTF from the hot tank to the cold tank can be maintained at constant flowrate and temperature, while the HTF from the cold tank to the hot tank varies with the insolation. Both the constant flowrate (m HTF ) and the variable flowrate (m c ) determine how much solar energy can be collected from the PTC and the efficiency of the ORC. In addition, the temperatures of the hot and cold tanks are critical for the integrated system for the following reasons: Both temperatures exert a great influence on the thermal efficiency of the ORC sub-system. The higher the hot tank temperature, the higher the thermal efficiency of the ORC sub-system. However, the temperature of the hot tank cannot be very high since the efficiency of the PTC will degrade at higher temperatures, which means less heat can be absorbed by the solar Fig. 1. Flowsheet of the basic ORC system with two-tank energy storage [12]. Fig. 2. Flowsheet of the recuperative ORC system with two-tank energy storage [12].
collector. Therefore, there is a trade-off between the PTC efficiency and the ORC efficiency. Moreover, the operating conditions of the ORC subsystem affect the HTF inlet temperature to the cold tank, and thus influences the PTC sub-system. It is obvious that the HTF is connecting the PTC, the thermal energy storage and the ORC sub-systems. The flowrate of the HTF, the target temperatures of hot and cold tanks, and the ORC operating conditions should be determined simultaneously to maximize the overall system efficiency of the solar energy power plant. The objective of this study is to determine the optimal design and operation of the integrated solar energy driven power plant, namely the optimal operating conditions of the ORC sub-system, the optimal design of the energy storage system, and the optimal operation of the PTC.

Modeling and optimization
To determine the optimal operation of the integrated solar-based power plant, Yang et al. [12] proposed a methodology for the optimal configuration of the system. The HTF inlet temperature to the evaporator is fixed at discrete values from 225 to 375 • C with 25 • C increments. For each HTF inlet temperature, an iterative procedure is performed to locate the optimal operating conditions of the integrated system. A large number of iterations is needed, and the process is tedious. Also, the optimal operating condition of the ORC system is determined under given HTF inlet temperatures, which means that the operating conditions of the ORC sub-system and the HTF inlet temperature are not optimized simultaneously. To address this shortcoming in their study, the HTF inlet temperature, the operating conditions of the ORC, and the optimal control of the energy storage system are optimized simultaneously in this work. An integrated model is developed in Matlab and Aspen HYSYS, which is a widely used process simulator, to obtain the optimal process design and control strategy of the solar energy driven ORC power plant. The thermal energy storage sub-system and the PTC sub-system are modeled in Matlab, while the ORC sub-system is simulated in Aspen HYSYS. The modeling of each sub-system is presented in the following.
The time horizon in this study is assumed to be one day as done by Yang et al. [12]. The direct normal irradiance (DNI) is divided into 24 time-intervals for one day. The irradiation and ambient temperature are assumed to be constant for each time-interval. The available energy from the sun in each time interval, Q s , can be calculated by Eq. (1).
where A c is the aperture area of the collector and G b is the beam irradiation or the DNI. The efficiency of the collector can be defined by Eq. (2).
where Q c is the energy absorbed by the HTF in the collector. The PTC chosen for modeling purposes in this study is the commercially available EuroTrough ET-150. This model was found to be both economical and effective for similar system layouts by Tzivanidis et al. [13]. The efficiency of the collector is given by Eq. (3), as suggested by Blanco et al. [17].
where G b is the beam irradiation and ΔT is the difference between ambient temperature and mean temperature in the solar collector, as shown in Eq. (4) [17].
where T amb is the ambient temperature, which also has an impact on the system.
Given the temperatures of the HTF at the inlet and outlet of the collector and the mass flowrate through the collector, the heat absorbed by the HTF in the solar collector can be calculated by Eq. (5).
where T hot and T cold are the temperatures of the hot and cold tanks, and C HTF p is the specific heat capacity of the HTF. Synthetic organic thermal oils are commonly used as the HTF in parabolic trough collectors. A mixture of Diphenyl Oxide and Biphenyl, with mass fractions of 73.5% and 26.5% respectively, is chosen as the HTF because the mixture can be used for temperatures up to 400 • C [18].
To determine the mass flowrate of the HTF, the specific heat capacity should be given. However, the specific heat capacity of the thermal oil mixture is unknown, and it is a function of temperature. Therefore, the specific heat capacities under different temperatures are retrieved from multiple simulations in Aspen HYSYS. Based on the simulation results, the specific heat capacity of the HTF is regressed as a function of temperature and shown in Eq. (6).
Then the mass flowrate of the thermal oil to the ORC sub-system can be calculated by Eq. (7).
where i denotes the time intervals, which are assumed to be 24 in this study.
The evaporator of the ORC sub-system connects the energy storage system with the ORC sub-system. The energy balance between the organic working fluid in the ORC and the HTF in the solar energy storage system can be expressed by Eq. (8).
where m ORC denotes the mass flowrate of the organic working fluid in the ORC system, and h 2 and h 3 are the specific enthalpies of the working fluid at state points 2 and 3 respectively, as shown in Fig. 1. The values of these parameters can be obtained from the Aspen HYSYS simulator. Aspen HYSYS is interfaced with Matlab with the help of Actxserver. For the recuperative ORC system, E. (8) is modified to Eq. (9).
The ORC model is built in Aspen HYSYS, and the Peng-Robinson equation of state is chosen as the thermodynamic property method [19,20], which has been widely used to simulate ORC systems [21,22]. To be consistent with the work of Yang et al. [12], the same assumptions are adopted: (i) the solar irradiation and ambient temperature for each hour are assumed to be constant; (ii) pressure drop and heat loss for heat exchangers and storage tanks are neglected; (iii) condensation temperature of the ORC is assumed to be 50 • C; (iv) minimum heat transfer approach temperature is set to 20 • C; and (v) isentropic efficiencies of the turbine and pump are assumed to be 80% and 75% respectively.
The net power output of the ORC sub-system is calculated by Eq. (10).
The thermal efficiency of the ORC sub-system is defined by Eq. (11).
The overall system efficiency is defined by Eq. (12).
The objective function is the overall system efficiency, which is equivalent to maximizing the power output of the ORC system for a given area of solar collectors and DNI values.
Based on the ORC model in Aspen HYSYS, the collector model and thermal energy storage model in Matlab, the integrated system can be simulated and evaluated. The optimal operating conditions of the ORC sub-system, the optimal design of the energy storage system and the optimal operation of the solar collector system should be determined simultaneously by optimization. The particle swarm optimization (PSO) algorithm [23], inspired by the behavior of the flock and their ability to localize food as a group, is a computationally efficient algorithm and chosen for optimization in this study. The PSO algorithm is a metaheuristic algorithm, which does not require gradient information and can escape from local optima. In addition, the PSO algorithm has the advantages of few tuning parameters and ease of implementation [24]. A population (called swarm) of potential solutions (called particles) moves around in the search space according to the particles positions and velocities [25]. The position of each particle in the PSO stands for a potential solution in the searching space. Each particle's movement is influenced by both its best-known position and the swarm's best-known position. This can speed up the process of locating the optimal solution [26]. This algorithm has been applied to many engineering problems and the effectiveness has been proven. Garg and Orosz [27] performed a thermo-economic optimization of a one-tank ORC system for waste heat and solar applications with the PSO algorithm. Liu et al. [28] applied the PSO algorithm to the multi-objective optimization of the fin-and-tube evaporator in a diesel engine ORC system. The PSO algorithm has been demonstrated to be robust in non-linear programming such as in heat exchanger network synthesis as well [29].
The PSO algorithm combining with Aspen HYSYS and Matlab capability is adopted to optimize the integrated ORC system in this study. The flowchart of the simulation-based optimization framework is illustrated in Fig. 3. This framework takes advantage of the modeling capability of Aspen HYSYS and Matlab, and the optimization capability of the PSO algorithm. The PSO algorithm generates the initial values of the decision variables based on the given upper and lower bounds. After the initialization, the decision variables are sent to the ORC model in Aspen HYSYS and the solar collector and energy storage system model in Matlab. However, the initial values may be infeasible. If the models do not converge, a large enough number is returned to the fitness function to penalize the infeasibility of the initial values. This is inspired by the penalty method for constrained optimization algorithms. If the models converge, the algorithm proceeds to the next step, where the constraints are evaluated. If the constraints are violated, a larger number is returned to the fitness function. After checking the constraints, the fitness function is evaluated for all the particles and the best particle can be obtained. Finally, the stop criteria of the algorithm are checked. If none of the criteria are satisfied, the algorithm proceeds to the next iteration, and the velocity and position of each particle will be updated based on information from the previous iterations. The optimal results are obtained when one of the stop criteria is satisfied.
Based on a degree freedom analysis for the basic ORC system, the following 4 variables are chosen as the independent variables: (1) hot tank temperature; (2) old tank temperature; (3) mass flowrate of the organic working fluid of the ORC; and (4) evaporation pressure of the ORC. For the recuperative ORC system, other than the above 4 degree of freedom, the heat load of the recuperator is chosen as the fifth independent variable since the heat exchanger heat load is easy to transfer between Matlab and Aspen HYSYS compared with other parameters in the model. The lower and upper bounds of the independent variables are listed in Table 2. It should be noted that the maximum evaporation pressure is set as 90% of the working fluid critical pressure. The condensation temperature, collectors' area, DNI, ambient temperature, turbine isentropic efficiency, pump isentropic efficiency, and motor efficiency are given the same values as used in Yang et al.'s work [12]. Since we will compare our results with Yang et al.'s work, the same assumptions and parameters are used in this study. If other assumptions are used, the results will be different, but the methodology is still valid.
The following constraints are considered in this study: (1) minimum approach temperature of the recuperator and evaporator must be greater than 3 • C to avoid too large heat exchanger; (2) vapor fraction of the turbine inlet stream must be 100%; (3) vapor fraction of the turbine outlet stream must greater or equal to 95% to avoid blade erosion due to droplets formed at the outlet of the turbine; and (4) vapor fraction of the pump inlet stream must be 0 to guarantee normal operation of the pump. Finally, the optimization model can be formulated as follows:  To make sure the search space is within the feasible region, a penalty function method is adopted to handle these constraints. Once a constraint is violated, a large penalty is added to the objective function to counteract the violation of the constraint, as shown in the flowchart in Fig. 3. The algorithm terminates when the maximum number of iterations or a specified tolerance is reached. The parameters of the PSO algorithm are set as follows: The maximum number of iterations is 100, the swarm size is 50, and the stop tolerance is set as 1e− 5. The integrated optimization model is solved on a PC with 4 cores 2.8 GHz Intel i7 CUP and 32 GB of RAM running Windows 10.

Results and discussion
The DNI value and the corresponding ambient temperature profile are given and illustrated in Fig. 4. The DNI values taken from Yang et al. [12] represent a typical summer day in Yinchuan (38 • 28 ′ 59 ′ 'N, 106 • 13 ′ 1 ′ 'E), in the northwest part of China. Four working fluids (toluene, cyclohexane, hexamethyldisiloxane (HMDSO) and n-pentane) studied by Yang et al. [12] and two additional working fluids (benzene and n-hexane) are investigated in this study. The critical parameters of the investigated working fluids are listed in Table 3. Since the condensation temperature is assumed to be 50 • C, the saturation pressure at 50 • C indicates the condensation pressure. The saturation temperature at 1 bar is also an important parameter, because 1 bar is the minimum condensation pressure if vacuum operation is not allowed in the system. As shown in Table 3, n-pentane has the lowest condensation temperature at 1 bar, while other working fluids have considerably higher condensation temperatures. A high condensation temperature indicates a low thermal efficiency of the ORC.

Basic ORC versus recuperative ORC
The optimal results of the basic ORC and the recuperative ORC are listed in Table 4. It is clear that for all investigated working fluids, the recuperative ORC can improve both the ORC thermal efficiency and the system efficiency significantly. For all of the working fluids investigated in this study, the improvement in ORC thermal efficiency lies in the range between 11.2 and 18.7% points, while the improvement in overall system efficiency is between 6.9 and 12% points. Toluene has the maximum power output for both basic and recuperative ORCs. The ORC thermal efficiency is improved from 24.3% to 36.3% with toluene as the working fluid. The overall system efficiency is improved from 17.4% to 24.8%. This shows that the recuperator can improve ORC thermal efficiency and overall system efficiency substantially. For the basic ORC, toluene and benzene perform much better than the other working fluids studied, while the superiority becomes marginal in the recuperative ORC. Hexamethyldisiloxane achieves the largest improvement when using recuperative ORC (18.7 and 12.0% points for ORC thermal efficiency and overall system efficiency respectively). For ORC applications in waste heat recovery, the recuperator is not always beneficial [30]. However, the results show that for the solar energy driven ORC system in this study, a recuperator should be configured.
Yang et al. [12] concluded that toluene has the best performance, which agrees well with our results in Table 4. However, the optimal system design derived from our work performs much better than the results reported by Yang et al. [12]. The reason is that our methodology optimizes the solar collector, the energy storage system and the ORC system simultaneously, and the optimal trade-off between the ORC system efficiency and the solar collector efficiency can be determined automatically by the PSO algorithm. Table 5 presents a detailed comparison between the results by Yang et al. [12] and the results obtained in this study.
In Yang et al. [12], the hot tank temperature is 375 • C, which is the highest temperature among all tested temperatures. However, for the basic ORC in our study, the optimal hot tank temperature is 368 • C, which is found by the optimizer. The cold tank temperature in our study is also lower than the reported value. For the basic ORC, the thermal efficiency is improved from 22.2% to 24.3%, while the system efficiency Fig. 4. The DNI and ambient temperatures for the given day [12].  Fig. 5. The flat range in working fluid curves for both basic ORC and recuperative ORC denotes a phase change. Since the optimal evaporation pressure is located at the upper bound (90% of the critical pressure of the working fluid), the flat range does not take up a big portion of the whole curve. It can be seen that the degree of superheating at the turbine inlet is very high for the recuperative ORC, thus the HTF and the toluene match well and the system efficiency is higher for the recuperative ORC (see Fig. 5).
The evaporation pressure in the basic ORC and recuperative ORC reaches the upper bound (37.12 bar) in both our study and the one by Yang et al. [12], which indicates that both the basic ORC and the recuperative ORC favor high evaporation pressure. In this study, the maximum evaporation pressure is set to 90% of the critical pressure to guarantee stable operation and reliable simulation results for the subcritical ORC. The upper bound can be relaxed if a supercritical ORC is considered, which will be discussed in detail later.
For the subcritical ORC without superheating, the pinch point between the working fluid and the heat source can be located either at the starting point of vaporization or at the starting point of preheating, also referred to as Vaporization Pinch Point (VPP) and Preheating Pinch Point (PPP) [31]. With reference to the results of the recuperative ORC with superheating as shown in Fig. 5, the pinch point can also be located at the superheating end, which can be termed Superheating Pinch Point (SPP). However, the degree of superheating in recuperative ORC can is around 100 • C as shown in Fig. 5. A large degree of superheating can increase the capital cost of the evaporator. To avoid too large degree of superheating, an easy way to handle this problem is to add a constraint to the model Since the HTF is a mixture, and the temperature profile is not a straight line, the pinch point can also be located in the preheating process as shown in Fig. 5 for the basic ORC. In both the basic and the recuperative ORCs, the phase change process exhibits a large temperature difference, which results in exergy losses.
Another considerable difference between the basic and recuperative ORCs is the cold tank temperature. As already mentioned with reference to Table 5, the cold tank temperatures are 57.6 • C and 234.3 • C for the basic and recuperative ORCs, respectively. This significant improvement in cold tank temperature can be explained as follows: The recuperator can preheat the working fluid to a higher temperature, which drives the cold tank temperature to a higher level. Therefore, both the ORC thermal efficiency and the system efficiency are improved with a recuperator.

Impact of working fluids
To investigate the effect of working fluids on system performance, other than the four working fluids studied by Yang et al. [12], two more working fluids (benzene and n-hexane) are investigated in this study. While toluene has the highest ORC thermal efficiency and system efficiency, one disadvantage is that its condensation pressure is less than the ambient pressure. This means that the turbine outlet stream is in vacuum, which can result in operational difficulties and safety issues for the system. Generally speaking, it is desired that the condensation pressure of the ORC is above ambient pressure. However, among all the investigated working fluids, only n-pentane can avoid vacuum condensation. The last column of Table 3 lists the saturation pressures at 50 • C. If the other working fluids also operate above ambient pressure, the condensation temperatures will be much higher than 50 • C, and the exact values are listed in the second last column of Table 3. If vacuum condensation is not allowed, the efficiency of the other working fluids will be significantly decreased. This means that n-pentane performs much better than the other working fluids if vacuum is not allowed in the ORC. In addition, for the recuperative ORC, the performance of npentane is not too far behind the other working fluids. From a practical point of view, n-pentane could therefore be the best choice.
Based on the results from this study, it can be concluded that the working fluids with higher critical temperature tend to have higher thermal efficiency, but the condensation pressure can be lower than the ambient pressure. Vacuum condensation will increase the operation  Table 5 Comparison with Yang et al. [12] using toluene as the working fluid.
Cycle type Yang et al. [ complexity, and there is a trade-off between thermal efficiency and operational complexity. The working fluids featuring both higher critical pressure and higher saturation pressure for the condensation temperature are desired for the ORC system. If this issue is considered, computer-aided molecular design techniques may discover more promising working fluids for the ORC.

Subcritical versus supercritical ORC
As already mentioned, the optimal evaporation pressure reaches the upper bound for both the basic and recuperative ORC, which indicates that the evaporation pressure could be a bottleneck in the system. To investigate the impact of the evaporation pressure on the system performance, the upper bound of the evaporator pressure is relaxed to supercritical pressures. With n-hexane as the working fluid, the upper bound of the evaporation pressure is set to 90% of the critical pressure (30.31 bar, see Table 3) for subcritical ORC and 100 bar for supercritical ORC respectively. When the upper bound of the evaporation pressure is less than the critical pressure, there is always phase change in the evaporator. In contrast, the ORC can be either subcritical or supercritical if the upper bound of the evaporation pressure is relaxed to 100 bar.
The last row in Table 4 lists the optimal results for the recuperative supercritical ORC with n-pentane as the working fluid. Compared with the subcritical ORC, the net power output is increased from 85.95 kW to 95.31 kW. The ORC thermal efficiency is improved by 11.3% (from 31.0% to 34.5%), and the system efficiency is improved by 10.8% (from 21.2% to 23.5%). Fig. 6 illustrates the temperature profiles in the recuperator and evaporator under both subcritical and supercritical  conditions. Interestingly, for the subcritical ORC, the phase change of npentane takes place in the recuperator, while the evaporator just superheats the working fluid. The phase change causes a flat range for the working fluid curve in the subcritical ORC. On the contrary, there is no such flat range for the working fluid curve in the supercritical ORC. The working fluid and the HTF match very well in the evaporator. The temperature profiles in the evaporator are almost the same for the subcritical and supercritical ORC systems, while the temperature profiles are quite different in the recuperator.

Optimal operation of the solar collectors
The optimal operation of this system refers to the control of the mass flowrate of the HTF to the solar collector. This is the only dynamic variable in the system, while other parts of the system should be in steady state. Only if the mass flowrate of the HTF is controlled properly, the system can be operated steadily at nominal design conditions. Once the cold and hot tank temperatures are determined, the mass flowrate of the HTF can be calculated by Eq. (5). Fig. 7 illustrates the optimal mass flowrate of the HTF in each time interval for a given day with toluene as the working fluid. The corresponding control measures can be taken to achieve the optimal operation of the integrated system. It is obvious that the mass flowrate of the HTF varies with the DNI value. For the recuperative ORC, the mass flowrate of the HTF is much higher than for the basic ORC. The mass flowrate of HTF in the recuperative ORC is 55.5% higher than that of the basic ORC, which indicates the tank size in the recuperative ORC should be at least 1.55 times of the tanks in the basic ORC. This can be attributed to the small temperature difference at the solar collector inlet and outlet in the recuperative ORC. The recuperative ORC outperforms the basic ORC in terms of ORC thermal efficiency and overall system efficiency. However, the basic ORC has the advantage of lower mass flowrate of HTF, which means the volumes of the storages tank can be smaller. If the volumes of the tanks become large, with a corresponding high capital cost, the HTF must be chosen carefully. An HTF with higher specific heat capacity will reduce the need for larger storage tanks. Also, other thermodynamic properties of the HTF should be considered in the selection process. Fig. 7 also illustrates the solar collector efficiency variation with time. It is interesting that the solar collector efficiency in the basic ORC is higher than that in the recuperative ORC. This can be attributed to the smaller temperature difference between the average solar collector temperature and ambient temperature in the recuperative ORC. Unlike the mass flowrate of the HTF, the efficiency of the solar collector is quite stable. The efficiency is almost constant except for the time without solar radiation.
In summary, the system configuration, choice of working fluid and the operating conditions (subcritical vs. supercritical) exert great influence on the system performance. All these factors have to be taken into account simultaneously while designing such a solar driven PTC power plant. Although the analysis is conducted for given DNI values for one day, the proposed model can be applied to different days, months, and locations. The DNI values can be predicted from historical data, and a more robust system design can be obtained based on more comprehensive DNI values. Moreover, the ORC power station can also act as a peak shaving power plant, which means that the ORC only operates a few hours per day. The corresponding optimal system design and operation can be determined in the same way based on the model developed in this study. Only the number of operating hours needs to be changed in the model. In Eq. (7), the number of hours should be changed from 24 to a specific number. In Eq. (12), however, only the numerator needs to be changed (from 24 to a specific number), while the denominator remains the same, since it denotes the total solar energy in one day.

Recent developments and future directions
Recently, Eterafi et al. [32] also investigated the solar driven ORC system with stable output. Domestic hot water production is considered alone with the ORC system for power generation. The prominent role of thermal energy storage system is also examined. The solar collector is parabolic dish concentrator (PDC) instead of PTC used in our study. Aghaziarati and Aghdam [33] performed the thermoeconomic analysis of a solar energy driven combined system consisting of ORC, cascaded refrigeration and heating system. They also compared three different types of solar collectors (PTC, PDC and LFR). Therefore, many more different types of solar collectors should be investigate and compared in the future research. Also, the cogeneration system design could be an interesting direction since multi-products can enhance the opportunities for process integration. Last but not least, the stochastic and seasonal variation of long-term solar radiation and weather impacts on the assessment of solar energy capture and use should be addressed. In this regard, stochastic optimization for the design and selection of candidate working fluids [34] under uncertainty [35] in the solar radiation will be investigated.

Conclusions
This study investigates the optimal design and operation of a solar energy driven ORC system with a parabolic trough collector and a twotank sensible thermal energy storage system. The energy storage system and the ORC system have been optimized simultaneously to achieve the best performance of the total system. The power cycle system can reach a stable and continuous operation through a given day with the help of the two-tank energy storage system. The optimal trade-off between the solar collector efficiency and the power output of the ORC system has been determined by the proposed simulation-based optimization framework. The optimal flowrate of the heat transfer fluid during the whole day can be determined based on this model. The impact of the ORC system configuration (with or without recuperator), working fluid selection and the mode of operation (i.e. subcritical vs. supercritical) on the system performance is analyzed. The thermal efficiency of recuperative ORC is about 11.2-18.7% points higher than that of a basic ORC system, while the improvement in overall system efficiency is about 6.9-12% points higher. Toluene performs best among all the investigated working fluids ignoring the problems with vacuum condensation. Compared with the study by Yang et al. [35], the overall system performance is improved substantially. Our methodology can improve the ORC thermal efficiency from 22.2% (previously reported value) to 24.3%, while the system efficiency is improved from 14.9% (previously reported value) to 17.4% for basic ORC. For the recuperative ORC, the ORC thermal efficiency is improved from 30.4% to 36.3%, while the overall system efficiency is improved from 17.9% to 24.8%. Using npentane as working fluid, vacuum condensation can be avoided, and both subcritical and supercritical operations are investigated with npentane as the working fluid. The supercritical ORC with n-pentane as working fluid can improve the ORC thermal efficiency by 11.3% and the system efficiency by 10.8% compared with the subcritical ORC. In this study, only sensible thermal energy storage is considered, while latent thermal energy storage systems or combined latent and sensible thermal energy storage systems are also interesting options, which deserve thorough comparative research in future work.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.